In electronic devices such as computers and portable devices, the graphics processing unit (GPU) is always one of the components that consume the most power. Accordingly, it is a very crucial topic to reduce power consumption to increase the service life of the battery for better power management. In a conventional electronic device, the GPU is usually controlled by software. For example, the central processing unit (CPU) is utilized to retrieve system information and user preferences and make a determination based on statistical information. The power management unit of the GPU in a conventional electronic device may turn off or turn on the associated partitions in the GPU, such as partitions allowing powering off, based on different conditions.
However, the granularity for turning on/off the partitions in the GPU by software is very large. That is, the power gating of the partitions having a fine granularity cannot be controlled by software. In addition, the CPU has to execute a corresponding application for monitoring and determining the power status. This may cause power consumption, and the performance of the CPU is also a bottleneck. Accordingly, there is a demand for a graphics processing system to solve the aforementioned power-management issues.